Market Shocks and Newspaper Ideology: Evidence from Taiwan

Home , Apple Daily, Apple Daily (Taiwan), Liberty Times

Market Shocks and Newspaper Ideology: Evidence from

Taiwan

Chun-Fang Chiang Long-Yu Chiou David Jinkins

Abstract

This paper investigates the influence of market forces on newspaper ideologies and readers’ choices in Taiwan. We measure a newspaper’s political position by counting a newspaper’s relative frequency of use of the terms “China” and “Mainland”. This captures a newspaper’s position on Taiwanese Independence, a major issue in Taiwanese politics. We find that Liberty Times had significantly changed its newspaper political position in 2003, when a tabloid newspaper owned by a Hong-Kong media firm entered the newspaper market in Taiwan. Using survey data, we estimate consumer preferences for newspaper consumption and examine the newspapers’ best strategies on choosing political positions. We find that the ideological change made by Liberty Times was necessary to maintain its market share in the face of competition. Finally, we analyze changes in newspaper ideologies in 2008 when China Times was sold to Want Want Group.

------

Chiang (corresponding author): Department of Economics, National Taiwan University, Taipei, Taiwan. TEL: 866-2-23519641; E-mail:[email protected]. Chiou: Department of Economics, National Taiwan University, Taipei, Taiwan. Jinkins: Department of Economics, Pennsylvania State University, PA, USA. Suppose Eve hadn't bitten into that forbidden fruit. There wouldn't be evil or wrongdoing in the world, and of course no news either.

– The founder of Apple Daily

1. Introduction

The naughties were frightening for a traditional Taiwanese newspaperman. In 2003, a tabloid newspaper from Hong Kong burst into the staid Taiwanese newspaper market. Apple Daily quickly devoured market share from established newspapers. Newspapers reacted to the competition by changing dramatically. Respectable newspapers began to publish more pictures, color spreads and less editorials, as well as increasing the size of headlines (Zhou 2004, Zhang 2005).

A theoretical paper by Mullanaithan and Schleifer (2003) implies that to avoid price wars, media firms react to competition by increasing news slants. In this paper, we first test if three major Taiwanese newspapers altered their political slants in addition to cosmetic changes. Using a divisive Taiwanese political issue we construct an ideology index that measures the frequency of using “China” instead of “Mainland” or “Mainland China” in a newspaper. We find that indeed Liberty Times increased its frequency of use of the term “China” significantly after Apple's appearance, while United Daily and China Times did not change their choices of words significantly. We also find that readers of major newspapers became more politically segregated.

Next we use the dynamic Taiwanese newspaper market to estimate consumer preferences for newspapers and ask how media sources choose slant. An individual-level survey with information on newspaper consumption and political preferences helps us to avoid potential problems of demand estimation when using aggregate market share data. The significant change in Liberty Times help us to identify consumer preferences for newspaper ideology. Combining our ideology index along with the survey data around 2003, we use nested logit model to estimate consumer preferences. This exercise is similar to the demand estimation in Gentzkow and Shapiro (2010) on newspaper ideology, except that our detailed household-level data allows us to use a much simpler specification. Using the estimated parameters in the utility function, we examine strategies made by newspapers when facing a new competitor. Simulations of the Taiwanese newspaper market show that the ideological change made by Liberty Times was necessary to maintain its market share in the face of competition but two other newspapers were leaving money on the newsstand. If one of these two newspapers shifted its editorial slant, it could increase readership.

The literature on media bias has argued that media owners may choose an editorial position in line with personal political views, even if such a position reduces profits (Povich 1996, Lichter, et al. 1986, Weaver and Wilhoit 1996, Balan, DeGraba, and Wickelgren (2009)). We test whether ownership affects the ideology using the 2008 purchase of a major newspaper, China Times, by a group with a China-friendly political position. The regression results suggest that the change in the ideology measure of China Times was statistically significant but small after the ownership transition. This paper adds evidence to the recent empirical literature on economic incentives and political bias in the media (Gentzkow and Shapiro, 2010; Shapiro and Sinkinson 2012). Gentzkow and Shapiro (2010) find that readers prefer to read newspapers with ideologies closer to their own and the owner of a newspaper does not affect its ideological slant. Shapiro and Sinkinson (2012) focus on the strategies of media firms under competition and show that newspaper competition in the early 20th century caused newspapers in the U.S. to choose different party affiliations.

General theoretical work on product differentiation is much older (Hotelling, 1929; d’Aspremont, Gabzwicz and Tisse 1979). Early duopoly studies of exactly one quality dimension and one variety dimension found that firms will maximize differentiation in one dimension and minimize in the other (Economides 1989; Neven and Thisse 1989). Later studies generalized this result to duopolies with any number of attribute dimensions, some considering only variety or quality differentiation (Ansari, Economides et al. 1998; Irmen and Thisse 1998; Garella and Lambertini 1999), and others considering both types (Dos Santos Ferreira and Thisse 1996). These studies suggest that firms will choose to differentiate maximally in one attribute in order to avoid harsh price competition, but will choose minimum differentiation in all other attributes to get the largest possible share of the market. 1 Vogel (2008) creates a model in which firms differ in productivity and concludes that more productive firms will be more isolated.

The paper is organized as follows. Section 2 introduces background of the newspaper market in Taiwan, our newspaper ideology index, and the data for estimation. Section 3 presents the econometric model of demand estimation. Section 4 presents the estimation results. Section 5 examines newspaper strategies under competition based on simulations. Section 6 concludes.

2. Background and Data

Before the entry of Apple Daily in 2003, there were three major newspapers in Taiwan, United Daily, China Times, and Liberty Times. These three newspapers comprised 80 percent of the newspaper market share. As shown in Figure 1, Apple Daily successfully entered the newspaper market in Taiwan in 2003 and became the second largest newspaper by 2005.2 Apple Daily is owned by a subdivision of Next Media, which is based in Hong Kong and is known for introducing tabloid-style journalism into Hong Kong. On the one hand, Apple Daily in Taiwan has been criticized for its sensational news style and has caused concern regarding possible violations of private rights tabloid news features. On the other hand, Apple Daily differs from the existing newspapers in its ability of attracting readers’ eyes. It has larger headlines, more pictures/tables, and more soft news with practical information. Overall, Apple Daily was able to leverage foreign knowledge and technology into Taiwan and appear as a strong competitor to existing newspapers.

Given its success in the newspaper market, it is natural to ask, then, why didn’t Apple Daily or a similar newspaper enter the Taiwan market earlier? The answer lies in

1 The degree of product differentiation may depend on market size. If market size is very large, the unique location equilibrium of firms is that both firms agglomerate at the center (Anderson, Simon and Neven, 1991). However, as the market size is relatively small, there exist two equilibria: one is the agglomeration but which is unstable, and the other is dispersed equilibrium. Furthermore the distance of the firms’ location would increase as market becomes smaller (Chamorro-Rivas, 2000. Liang et at., 2005). 2 Newspaper readership is calculated from the data of AGB Nielson. The readership of a particular newspaper is defined as the number of readers who read the particular newspaper yesterday divided by the number of people who read one or more newspapers yesterday. government regulation. Although Taiwan’s newspaper prohibition ended in 1988, it was not until the late 1990’s that the Taiwanese government began to relax its publishing laws relating to foreign companies. The change in policy was largely due to Taiwan’s application to join the World Trade Organization, which was accepted in 19993 with Taiwan officially becoming a member on January 1, 2002. As part of Taiwan’s WTO agreement, Taiwan cannot restrict the news agency services, except to require the residence of an editor and publisher in Taiwan (WTO 2001). It should be noted that the first 500,000 trial issues of Apple Daily were actually distributed on March 20, 2002, just three months after Taiwan entered the WTO (Zhou 2005).

2.1 Political Positions of Newspapers

There have been two dominant political forces in Taiwan after the martial law regime. One is the pro-unification Pan-Blue coalition, comprised of the Kuomintang, the People First Party (PFP), and the New Party. The other is the pro-independence Pan-Green coalition, comprised of the Democratic Progressive Party (DPP) and the Taiwan Solidarity Union (TSU). The Pan-Blue coalition and the Pan-Green coalition have different policies regarding cross-strait politics and their differences are reflected in the language used by politicians.

Researchers have come up with various ways to measure media slant. Groseclose and Milyo (2005) compare the think tanks and the policy groups cited by Congress members and those cited by news media outlets, and derive a way to measure the ideology for several major media outlets.4 Gentzkow and Shapiro (2010) compare the language used by Congress members and newspapers to measure media slant of newspapers. In our study, we use the frequency of using “China” instead of “Mainland” or “Mainland China” in article titles to measure political ideologies of newspapers in Taiwan. This captures a newspaper’s position on cross-strait politics, a major issue in Taiwanese politics. We searched for the phrases “China”, “Mainland China” and

3 Geopolitics dictated that the People’s Republic of China had to become a member before the Republic of China. Although Taiwan had fulfilled all the conditions required for entry by 1999, it had to wait until after China’s entry in December, 2001 for formal membership. 4 The ideology of Congress members is measured by adjusted ADA scores constructed by Groseclose, Levitt, and Snyder (1999). In their findings, there exists a strong liberal trend in the average member of Congress during 1947-94. “Mainland” in the article titles, and excluded phrases such as “China Bank” because the term China in this case does not refer to the country. There are two newspaper archives, the Taiwan News Smart Web and the Central News Agency Archive. The Taiwan News Smart Web covers most of the news articles but did not include( the Liberty Times until 2003, and Apple Daily was not included in the archive until 2005. The Central News Agency Archive covers all major newspapers but does not collect all news articles. We therefore use the Central News Agency Archive to examine changes in newspaper ideologies due to the entry of Apple Daily, and use Taiwan News Smart Web to examine changes in newspaper ideologies due to the ownership switch of China Times at the end of 2008.

As shown in Figure 2, we find that indeed Liberty Times increased its frequency of use of the term “China” significantly after Apple's appearance, while United Daily and China Times did not significantly change their terminology choices. In addition, we have checked that the general editor of Liberty Times was not replaced from 1996 to 2012 in the past 15 years,5 thus the significant change in the ideology measure of Liberty Times is unlikely attributed to the editor’s personal ideology.

2.2 Other Newspaper Characteristics

Other than political ideologies, some additional features influence readers’ choices of newspapers: price, diversity of news articles, layout, and other factors. Facing competition from Apple Daily, the retail prices of all three traditional newspapers declined from $15 to $10 in 2003. All three papers also changed their layout, from a vertical format to a horizontal format, the same layout as used by Apple Daily. One of the most significant features of Apple Daily was its undisguised and bloodthirsty style. The headline news articles of Apple Daily are often crime news or scandals with colorful and vivid pictures. While the tabloid style was criticized, it might have affected the news selection of other newspapers. We therefore include the ratio of crime news for one newspaper’s characteristics as a proxy for the tendency to report sensational news.6 We also include the diversity index reported by Lee (2007) in our estimation, although she finds that news diversity in the three traditionally dominant newspapers after was

5 Wen-Hui Hu took over the position of general editor after February, 2013, and the ex- general editor Jin-Rong Chen was promoted to the position of vice president. 6 The Criminal news includes murder, violence, sex crimes and larceny. not statistically different from that before the launch of Apple Daily in the Taiwan market.

Table 2 presents newspaper characteristics of three major newspapers before and after Apple Daily.

2.3 Readership Composition

Information on individual newspaper consumption came from TEDS and TSCS in the following years: 1998, 2000, 2001, 2003, 2004, 2005, and 2008. Table 1 provides a timeline of the surveys we use in this paper. These surveys have information on newspapers they read most and their background information such as age, gender, education, and political party preference.

Since newspaper ideologies varied over time, readership composition in terms of political preferences may have changed after the entry of Apple Daily into Taiwan. We classified respondents into three groups based on their preferred political party. Respondents who support KMT, NP or PFP are classified as Pan-Blue supporters. Those who support DPP or TSU are classified as Pan-Green supporters. Those who neither belong to Pan-Blue nor Pan-Green are Independents. Figure 3 presents choices of newspapers over four major newspapers by political preferences before and after the entry of Apple Daily. As shown, with or without Apple Daily, the Pan-Blue readers prefer United Daily to China Times to Liberty Times. On the other end of the political spectrum, the Pan-Green supporters prefer Liberty Times to China Times to United Daily. The segregation pattern is stronger after the entry of Apple Daily. In terms of readership, China Times and United Daily both lost some readership after facing completion, especially among independents. The influence of the new tabloid newspaper on the readership numbers of Liberty Times is relatively smaller.

The changes in the political composition of newspaper readers indicate that readers may react to changes in political ideologies of newspapers. In the next section, we estimate consumer preferences over newspapers and then examine newspapers’ choices on ideologies.

3. Consumer Preferences and Newspaper Ideologies

Combining information on newspaper ideologies and readers’ choices of newspapers, we estimate consumer preferences by a two level Nested Logit model. The first level offers distinct choices into four types: reading a traditional newspaper, reading an entertainment newspaper, Apple Daily, reading other newspapers, and not reading a newspaper. In the second level, the traditional newspapers include Liberty Times, United Daily, and China Times, as presented in Figure 4. The nested logit model permits the choices of United Daily, Liberty Times, and China Times to be correlated.

In the market of traditional newspapers, readers can choose from Liberty Times, United Daily, and China Times. Other than reading a traditional newspaper, an individual may also choose to read Apple Daily when Apple Daily is available, other newspapers, or to not read a newspaper.

The random utility of reader i to read newspaper k in type j can be expressed as:

, (1)

, where includes newspaper characteristics, and represents individual

characteristics. The error terms are distributed Gumbel’s multivariate extreme-value distribution. We assume that readers care for political ideologies of newspapers only when reading a traditional newspaper. The utility of reader i when reading a traditional newspaper k can be expressed as:

, (2)

is the ideology of newspaper k, and is the preferred news slant of reader i. A

higher value of or denotes ideologies closer to the ideologies of Pan-Green Coalition. We assume that the readers’ preferred media slant, can be approximated by their political party preferences.

3 Substituting equation (3) into equation (2), we have

1,…,,1,…, Where 2, , 2, 2.

In our estimation, we use the variable fraction of China, the frequency of the word China” instead of “Mainland” or “Mainland China” in the article titles of the newspaper,

as the measure of newspaper ideology. X includes individual characteristics such as party preferences, age, gender, and years of schooling that would affect choices of newspapers. Z includes newspaper price, diversity index, and crime news ratio.

Without newspaper fixed effects, the coefficient of fraction of China^2 will be , and the coefficients of the interaction terms of fraction of China and political party preferences (Green and Blue) will be 2 and 2 respectively. We expect that derived β and a will be negative, and derived b to be positive, meaning that readers prefer to read a newspaper with their preferred news slant, and their preferred news slant is correlated with their political preferences.7

4. Empirical Strategy and Estimation Results

We are interested in estimating parameters in the model of newspaper demand. Without exogenous variation in newspaper ideologies, the correlation between newspaper ideologies and consumer ideologies could be due to reverse causality. Newspaper ideology may affect readers’ beliefs and therefore the correlation between newspaper ideologies and consumer ideologies could reflect newspapers’ influence instead of readers’ preferences. We use a survey sample from 2000 to 2005 to estimate parameters in the demand model. During the sample period, the change in newspaper ideologies was mainly driven by the entry of Apple Daily.

A. Results from the baseline model

We first estimate the model without newspaper × time fixed effects, and the results are reported in 3. The coefficient of newspaper ideologies is negative and significant, implying that readers do care about the ideologies of newspapers. The coefficients of the interaction terms, fraction of China*Blue and fraction of China*Green, are negative and positive respectively, implying that the readers’ preferred newspaper ideologies are correlated with their own ideologies. These results suggest that Pan-green supporters prefer to read newspapers with positions on pro independence, and Pan-blue supporters prefer to read newspapers with China friendly positions.

In Table 4A, besides newspaper ideology, we include other newspaper characteristics such as the ratio of crime news, news diversity index, and the retail price

7 All results in the paper are estimated by the software “Limdep-NLOGIT 5”. in our estimation. Table 4B presents partial/marginal effects based on the estimation results in Table 4A.

B. Results with newspaper × time fixed effects

While we have a ratio of crime news, the news diversity index, and retail prices, we may not capture all the newspaper characteristics that changed after the entry of Apple Daily. Therefore, in Table 5, we include newspaper fixed effects and their interaction terms with a dummy variable, After, which indicates that the respondent was interviewed after the entry of Apple Daily. These variables help us to capture the influence of unobserved newspaper characteristics on overall readership.

In Table 6, we include newspaper × month fixed effects instead of newspaper × After fixed effects. By doing so, the variables fraction of China and its square term are dropped in the estimation. The coefficient of fraction of China*Blue remains negative and significant, and the coefficient of fraction of China*Green remains positive and significant, suggesting that readers are more likely to read newspapers with ideologies approximating their own ideologies.

C. Robustness Checks

The people who support Pan-Blue coalition and the people who support Pan-Green coalition usually have different characteristics. The Pan-Blue coalition supporters are usually more educated and have higher income. While the changes made by newspapers could have also changed readership composition by education or income, we include the interaction terms of newspaper ideologies and the readers’ years of schooling and annual income in our estimation as robustness checks. The results are reported in Table 7 columns (1) to (3).

We also restrict the samples in 2001 and 2003 to estimate consumer preferences again. On the one hand, the variation in newspaper ideologies will mainly came from dramatic ideological changes due to competition from Apple Daily. On the other hand, the influence of the newspaper on readers’ beliefs is unlikely to have large effects in a short period of time. These estimation results are presented in the last column of Table 7. Overall, the results are consistent with previous findings.

5. Would a newspaper gain more readership by changing its political ideologies?

It is possible that not all newspapers have selected their optimal positions to maximize their profits. For example, newspaper ideologies may be affected by preferences of the owner, editors, and reporters. Facing the entry of Apple Daily, we observed that Liberty Times responded by changing its position significantly while United Daily did not adjust its political position. Here we examine whether these newspapers had made the right choices to maintain their market share. Using consumer preferences estimated above, we predict their market share if the newspapers had not changed their political ideologies after the entry of Apple Daily. The results are presented in Table 8. The simulated results suggest that the significant change in political positions made by Liberty Times had increased its market share by 10% in the competition.

We also further explore the possibility for newspapers to gain more market share by changing their political ideologies. First we simulate the market share of Liberty Times under different political ideologies without changing ideologies of the other newspapers. As shown in Table 9A, there is not much room for Liberty Times to further increase its market share. The ideology, measured by fraction of China, of Liberty Times after Apple Daily was 0.95, and decreasing the value will not give Liberty Times more readers. However, the results in Table 9B indicate that China Times and United Daily could have gained more market share by adjusting their political ideologies.

Finally we calculate the newspaper ideologies in the Nash Equilibrium, given the condition that the ideology of Liberty Times is 1. As shown in Table 10, in the Nash Equilibrium, the ideology of China Times is 0.471, and the ideology of United Daily is 0.398. The market shares of China Times and United Daily are larger than their actual/predicted market share under the observed ideologies. These results further confirmed that China Times and United Daily could have gained more market share by adjusting their political ideologies.

6. Ownership Transition

As the discussion in the literature, a newspaper owner’s political ideologies may affect the news slants of newspapers. We are interested in whether the ideology of China Times slants toward China after having been sold to Want Want group which is a pro-China enterprise in November 2008. We continue to use the ratio of China mentions as a measurement. Figure 6 shows that the ideology of China Times was closer to that of the United Daily after it was sold to Want Want group. To examine if the change in ideologies are statistically significant, we use newspaper ideology as a dependent variable in OLS regression. In the table 9, the estimation results suggest that the frequency of using the word “China” had decreased approximately 10 % around the beginning of 2009.

7. Conclusion In this paper, we investigate the influence of market forces and owner preferences on newspaper ideologies and readers’ choices from 2002 to 2009 in Taiwan. We measure a newspaper’s political position by counting a newspaper’s relative frequency of use of the terms “China” and “Mainland”. This captures a newspaper’s position on Taiwanese Independence, a major issue in Taiwanese politics. We find that newspapers significantly changed their newspaper political positions in 2003, when a tabloid newspaper owned by a Hong-Kong media firm entered the newspaper market in Taiwan. Using survey data, we estimate consumers’ preferences of newspaper reading and find that readers are sensitive to newspaper ideologies. Then we examine newspapers’ best strategies on choosing political positions in the face of competition. We find that the ideological change made by Liberty Times was necessary to maintain its market share in the face of competition, and China Times and United Daily could have gained more market share by adjusting their political ideologies.

References

Ansari, A., N. Economides, et al. (1998). "The Max-Min-Min Principle of Product Differentiation." Journal of Regional Science 38(2): 207-230. Brenner, S. H. (2005). "Hotelling Games with Three, Four, and More Players*." Journal of Regional Science 45(4): 851-864. d'Aspremont, C., J. J. Gabszewicz, et al. (1979). "On Hotelling's "Stability in Competition"." Econometrica 47(5): 1145-1150. Dos Santos Ferreira, R. and J.-F. Thisse (1996). "Horizontal and vertical differentiation: The Launhardt model." International Journal of Industrial Organization 14(4): 485-506. Economides, N. (1989). "Quality variations and maximal variety differentiation." Regional Science and Urban Economics 19(1): 21-29. Garella, P. G. and L. Lambertini (1999). "Good vs. bad characteristics in vertical differentiation." Economics Letters 65(2): 245-248. Gentzkow, Matthew and Jesse M. Shapiro 2010. "What drives media slant? Evidence from US daily newspapers." Econometrica 78(1): 35-71. Gentzkow, M., J. M. Shapiro and M. Sinkinson, 2010. "Competition and Ideological Diversity: Historical Evidence from US Newspapers." Working paper. Groseclose, Tim and Jeffrey Milyo 2005. "A measure of media bias." Quarterly Journal of Economics 120(4): 1191-1237. Groseclose, Tim, Steven D. Levitt, and James M. Snyder, Jr., 1999. "Comparing Interest Group Scores across Time and Chambers: Adjusted ADA Scores for the U. S. Congress." American Political Science Review XCIII (1999), 33–50. Vogel, J, 2008. " Spatial Competition with Heterogeneous Firms." Journal of Political Economy, 116(3): 423-466. Ho, S. J. and M. P. Sun (2008). "Hetereogeneous Multi-Product Competition in Taiwan's Newspaper Industry." Asian Journal of Communication 18(2): 102-116. Irmen, A. and J.-F. Thisse (1998). "Competition in Multi-characteristics Spaces: Hotelling Was Almost Right." Journal of Economic Theory 78(1): 76-102. Lee, Y.-C. (2007). "Effects of Market Competition on Taiwan Newspaper Diversity." Journal of Media Economics 20(2): 139 - 156. Mazzeo, M. J. (2002). "Product Choice and Oligopoly Market Structure." The RAND Journal of Economics 33(2): 221-242. MOJ, M. o. J. (2009). "National Law and Regulation Database (全國法規資料庫)." Retrieved July 8, 2009.. Seim, K. (2006). "An Empirical Model of Firm Entry with Endogenous Product-Type Choices." The RAND Journal of Economics 37(3): 619-640. Shaked, A. and J. Sutton (1982). "Relaxing Price Competition Through Product Differentiation." The Review of Economic Studies 49(1): 3-13. WTO (2001). Schedule of Specific Commitments on Services List of Article II MFN Exemptions. WT/ACC/TPKM/18/Add.2. WTO. Zhang, H.-. (2004). Apple Daily's Influence on United Daily and China Times (蘋果日報對聯合報，中國時報的影響). Graduate Institute of News, National Cheng Chi University. MA: 122. Zhou, S. (2005). One becomes what one studies: The meaning of layout design competition--the example of China Times, United Daily, and Freedom Times before and after the publishing of Apple Daily (學什麽？像什麽：報紙版面設計之競爭意涵－以蘋果日報出刊前後中時聯合自由三包為例). Graduate Institute of Broadcasting, National Chong Cheng University . MA: 186.

Normalized Newspaper Readership .4 .3 .2 Percentage .1

1997Q1 1999Q1 2001Q1 2003Q1 2005Q1 2007Q1 2009Q1 2011Q1 Time line China Times Liberty Times United Daily Apple Daily Other newspapers

Figure 1: Newspaper readership 1 .8 .6 .4 Fraction of China .2 0

1998-01 2000-01 2002-01 2004-01 2006-01 2008-01 Date

China Times United Daily Liberty Times

Figure 2: Newspaper ideologies Readership Readership Before entry of Apple Daily After entry of Apple Daily .3 .2 percentage .1 0 blue no party green blue no party green Source: sample from TEDS and TSCS Source: sample from TEDS and TSCS

United China times Liberty Apple

Graphs by apple_entry

Figure 3: Newspaper Reading by Political Preferences

Readers

Other General Not reading Apple Daily Newspapers newspapers newspapers

United Daily China times Liberty Times

Figure 4. Choice Tree in the nested logit model

Prediced Probability Prediced Probability Before Apple Daily After Apple Daily .4 .3 .2 Percentage .1 0 Bule Neutral Green Bule Neutral Green Source: sample from TEDS and TSCS Source: sample from TEDS and TSCS

United Dialy China times Liberty Times

Figure 5. Predicted newspaper reading by political preferences 1 .8 .6 Fraction of China .4 .2

2008-01 2008-07 2009-01 2009-07 2010-01 Date China Times United Daily Liberty Times

Figure 6 Newspaper ideologies around the end of 2008

Table 1: TEDS and TSCS Survey Outline

Survey Date Sample size

TSCS 1998.07.05-08.18 1,435

TSCS 2000.07.22-08.03 1,960

TEDS 2002.01.04-04.02 2,022

TSCS 2003.08.10-12.01 2,158

TEDS 2004.01.15-09.31 3,075

TSCS 2005.07.05-09.18 2,146

TEDS 2008.01.10-09.05 2,945

Table 2. The Newspapers characteristics (2000 - 2005) China Times United Daily Liberty Times Before After Change Before After Change Before After Change Fraction of China 0.178 0.229 28.65% 0.164 0.171 4.27% 0.602 0.948 57.48% Diversity index 0.857 0.862 0.58% 0.817 0.857 4.90% 0.841 0.855 1.66% Crime news ratio 5.559 5.963 7.27% 6.961 5.601 -19.54% 4.872 6.110 25.41% Retailing Price $15 $10 -33.33% $15 $10 -33.33% $10 $10 0%

Table 3. Consumer Preferences A. Newspaper characteristics 2.38*** Fraction of China (0.83) -0.61** Fraction of China^2 (0.31) -0.94*** Blue*Fraction of China (0.44) 0.48*** Green*Fraction of China (0.22) Newspapers CT UD LT AD Others B. Individual characteristics 0.81*** 0.94*** 0.30 1.18*** 0.46*** Pan-Blue (0.13) (0.12) (0.33) (0.30) (0.09) 0.48 -0.51*** 0.82** 0.05 0.27*** Pan-Green (0.38) (0.12) (0.37) (0.27) (0.09) 0.05 0.14* 0.08 0.04 -0.01 Sex (0.08) (0.08) (0.08) (0.11) (0.07) 0.04** 0.07*** 0.05*** 0.00 0.04*** Age (0.02) (0.02) (0.01) (0.02) (0.01) -0.00*** -0.00*** -0.00*** -0.00*** -0.00*** Age2 (0.00) (0.00) (0.00) (0.00) (0.00) 0.21*** 0.23*** 0.18*** 0.15*** 0.14*** Years of education (0.01) (0.01) (0.01) (0.02) (0.01) 0.39*** 0.25*** 0.26*** 0.24*** 0.17*** Log income (0.05) (0.05) (0.05) (0.07) (0.05) Note. 1.Sample: 2000-2005. 2. Coefficients reported. 3.Reference Choice: not to read a newspaper. 2. The number of observations is 7893.

Table 4A. Estimation results with more newspaper characteristics A. Newspaper characteristics 2.58** Fraction of China (1.11) -0.31*** Fraction of China^2 (0.05) -0.93** Blue*Fraction of China (0.42) 0.89** Green*Fraction of China (0.43) 2.21** Crime news ratio (1.09) 6.02** Diversity index (2.50) -0.11* Price (0.05) Newspapers CT UD LT AD Others B. Individual characteristics 0.79*** 0.94*** 0.29 1.17*** 0.46*** Pan-Blue (0.13) (0.12) (0.33) (0.30) (0.10) -0.08 -0.52*** 0.81** 0.04 0.28*** Pan-Green (0.12) (0.13) (0.37) (0.27) (0.09) 0.05 0.14* 0.08 0.05 -0.01 Sex (0.08) (0.08) (0.08) (0.11) (0.07) 0.04** 0.07*** 0.05*** 0.00 0.04*** Age (0.02) (0.02) (0.01) (0.03) (0.01) -0.00*** -0.00*** -0.00*** -0.00*** -0.00*** Age2 (0.00) (0.00) (0.00) (0.00) (0.00) 0.21*** 0.23*** 0.19*** 0.15*** 0.14*** Years of education (0.01) (0.01) (0.01) (0.02) (0.01) 0.39*** 0.25*** 0.26*** 0.23*** 0.17*** Log income (0.05) (0.05) (0.05) (0.08) (0.05) Note. 1.Sample: 2000-2005. 2. Coefficients reported. 3.Reference Choice: not to read a newspaper. 2.The number of observations is 7893.

Table 4B. Marginal Effects from Table 4A A. Newspaper characteristics Fraction of China 0.1622 Fraction of China^2 -0.0201 Blue*Fraction of China -0.0578 Green*Fraction of China 0.0568 Crime news ratio 0.1389 Diversity index 0.3759 Price 0.0069 Newspapers CT UD LT AD Others B. Individual characteristics Pan-Blue 0.0491 0.0593 0.0182 0.0735 0.0277 Pan-Green -0.0055 -0.0311 0.0512 0.0018 0.0176 Sex 0.0024 0.0088 0.0050 0.0018 -0.0001 Age 0.0025 0.0037 0.0031 0.0025 0.0025 Age2 0.0000 0.0000 0.0000 0.0000 0.0000 Years of Education 0.0131 0.0154 0.0132 0.0093 0.0088 Log-income 0.0245 0.0151 0.0163 0.0145 0.0107 Note. 1.Sample: 2000-2005. 2. Coefficients Reported. 3.Reference Choice: not to read a newspaper

Table 5. Consumer preference with newspaper * after fixed effects Newspaper characteristics 2.69*** Fraction of China (1.13) -0.48** Fraction of China^2 (0.23) -0.98** Blue*Fraction of China (0.43) 0.97** Green*Fraction of China (0.44)

After*China Times -1.74*** After*United Daily -1.48*** After*Liberty Times -2.37*** After*Others -2.31*** Newspapers CT UD LT AD Others Individual characteristics 0.82*** 0.95*** 0.15 1.38*** 0.37*** Blue (0.14) (0.13) (0.34) (0.31) (0.10) -0.22 -0.65*** 1.09*** -0.10 0.03 Green (0.13) (0.13) (0.38) (0.29) (0.10) 0.09 0.18** 0.11 0.08 0.04 Sex (0.08) (0.08) (0.07) (0.11) (0.08) 0.04*** 0.07*** 0.05*** 0.00 0.04*** Age (0.01) (0.01) (0.01) (0.02) (0.01) -0.00*** -0.00*** -0.00*** -0.00** -0.00** Age2 (0.00) (0.00) (0.00) (0.00) (0.00) 0.25*** 0.27*** 0.22*** 0.17*** 0.20*** Years of education (0.02) (0.02) (0.01) (0.02) (0.02) 0.42*** 0.29*** 0.29*** 0.22*** 0.22*** Log income (0.06) (0.05) (0.05) (0.08) (0.05) Newspaper*After fixed effects V V V V V Note. 1.Sample: 2000-2005. 2. Coefficients reported. 3.Reference Choice: not to read a newspaper. 2. The number of observations is 7893.

Table 5B. Marginal effect from table 5A. Newspaper characteristics Fraction of China 0.1712 Fraction of China^2 -0.0321 Blue*Fraction of China -0.0624 Green*Fraction of China 0.0617 After entry China Times -0.1107 United Daily -0.0942 Liberty Times -0.1510 Others -0.1471 Newspapers CT UD LT AD Others Individual characteristics Blue 0.0526 0.0615 0.0099 0.0864 0.0235 Green -0.0140 -0.0406 0.0694 -0.0064 0.0023 Sex 0.0057 0.0115 0.0078 0.0051 0.0025 Age 0.0025 0.0044 0.0032 0.0000 0.0025 Age2 0.0000 0.0000 0.0000 0.0000 0.0000 Years of education 0.0155 0.0172 0.0142 0.0108 0.0129 Log income 0.0248 0.0185 0.0185 0.0140 0.0140 Newspaper*After fixed effects V V V V V Note. 1.Sample: 2000-2005. 2. Coefficients reported. 3.Reference Choice: not to read a newspaper.

Table 6. Estimation with Newspaper*month fixed effects Newspaper characteristics -0.81*** Blue*Fraction of China (0.23) 0.89** Green*Fraction of China (0.40) Newspapers CT UD LT AD Others Individual characteristics 1.02*** 1.11*** 0.11 1.06*** 0.41*** Blue (0.12) (0.11) (0.34) (0.22) (0.07) -0.22 -0.64*** 1.06*** -0.19 0.13 Green (0.19) (0.17) (0.25) (0.30) (0.11) 0.07 0.14* 0.08 0.06 0.00 Sex (0.07) (0.06) (0.06) (0.14) (0.08) 0.04** 0.07*** 0.05*** -0.04 0.04*** Age (0.02) (0.01) (0.01) (0.03) (0.02) -0.00** -0.00*** -0.00*** -0.00 -0.00** Age2 (0.00) (0.00) (0.00) (0.00) (0.00) 0.29*** 0.29*** 0.24*** 0.12*** 0.21*** Years of education (0.02) (0.02) (0.01) (0.02) (0.03) 0.42*** 0.31*** 0.21*** -0.11* 0.20*** Log income (0.06) (0.04) (0.04) (0.04) (0.06) Newspapers*month V V V V V Note. 1.Sample: 2000-2005. 2. Coefficients reported. 3.Reference Choice: not to read a newspaper. 2. The number of observations is 7893.

Table 7. Robustness Checks Newspaper characteristics (1) (2) (3) (4) Sample:2001-2003 6.83*** 6.21*** 10.34*** 8.85*** 8.03*** Fraction of China (2.13) (1.91) (3.05) (2.78) (3.05) -1.88** -1.82** -2.97*** -3.17*** -3.57*** Fraction of China^2 (0.92) (0.88) (0.87) (0.81) (0.87) -0.96** 0.81** -0.88** -0.72*** Blue*Fraction of China (0.41) (0.39) (0.44) (0.24) 0.93*** 1.11*** 1.07** 0.57*** Green*Fraction of China (0.40) (0.41) (0.49) (0.19) Green index*Fraction of 0.27***

China (0.09) -0.02 -0.02 -0.07 -0.02 Taiwanese* Fraction of China (0.40) (0.40) (0.36) (0.40) Log-income* Fraction of -0.31 -0.31 -0.32 -0.31

China (0.25) (0.25) (0.23) (0.25) Education year* Fraction of -0.12*** -0.11** -0.11** -0.07

China (0.04) (0.05) (0.05) (0.05) No. of observations 7893 7893 7893 7893 3192 Newspaper fixed effects*After V V V V V Individual characteristics V V V V V

Table 8. Predicted and Simulated Market Share China Times United Daily Liberty Times Predicted Market Share after 22.42% 23.53% 32.66% Apple Daily Predicted Market Share before 25.82% 24.20% 26.41% Apple Daily Simulated Market Share if political positions unchanged in 22.06% 25.22% 23.68% the presence of Apple Daily. Note: Predictions and simulation are based on the estimation results of table 4A.

Table 9A. Simulated Market Share of LT after Apple Daily -5% -3% Baseline +3% Up to 100% Liberty Times 30.70% 31.22% 32.66% 33.61% 35.01% Note. The ideology index of Liberty Times was around 0.95 after the entry of Apple Daily.

Table 9B. Simulated Market Share after Apple Daily China Times United Daily Baseline 22.42% 23.53% +10% 24.90% 26.11% +20% 25.68% 27.51% +30% 26.42% 27.33% +40% 25.87% 24.92% +50% 21.54% 22.60% +60% 17.71% 19.04% +70% 13.90% 14.49%

Table 10. Newspaper Ideologies and Simulated Market Share under Nash Equilibrium China Times United Daily Liberty Times Observed Newspaper ideologies 0.23 0.17 0.95 Newspaper ideologies under 0.47 0.40 1 Nash equilibrium Market share under Nash 24.12% 25.67% 30.51% equilibrium Note: Simulation is based on the estimation results of table 4A.

Table 11. Ownership Transition and Newspaper Ideologies Dependent Variable: Fraction of China Sample Period 2000-2009 2003-2009 2008-2009 -0.08*** -0.11*** -0.13*** China Times*year2009 (0.03) (0.04) (0.05) 0.11*** 0.14*** 0.16*** China Times (0.02) (0.02) (0.04) 0.23*** 0.20*** 0.08*** Year2009 (0.02) (0.02) (0.03) 0.19*** 0.22*** 0.34*** Constant (0.01) (0.01) (0.03) Number of Observations 240 168 48 R squared 0.22 0.27 0.47 Note. Control group is United Daily.